An Improved SMAP Soil Moisture Retrieval using Deep Neural Network-based inversion of a Passive Microwave Radiative Transfer Model
Abstract
Soil moisture (SM) is an important variable affecting various hydrological processes. L-band brightness temperature (TB) measurement from satellite missions such as Soil Moisture Active Passive (SMAP) is widely used because of its sensitivity to surface SM variation (up to 5 cm). SM retrieval from the satellite measurement was widely conducted by a zero-th order approximation of the radiative transfer model (RTM), called τ-ω model. The components of τ-ω model consist of three parts those are: (1) direct upward emission from the bare soil, (2) direct upward emission from vegetation cover, and (3) reflected downward emission from vegetation cover by soil.
A large number of field or airborne experiments have been conducted to find relationships between each component of τ-ω model. Even those tough those extensive efforts, forested regions are less represented in the SM retrieval algorithm due to the limited amount of available data. In addition to, there is a problem of reduced sensitivity of L-band TB when increased vegetation water content (VWC). Consequently, SMAP SM retrievals are flagged as unreliable over dense forest regions (where VWC >= 5 kg/m2), and many studies have reported bias or performance degradation of SMAP SM over those regions. This study experimented deep neural network (DNN)-based inversion process of SMAP TB to overcome the aforementioned vegetation-related errors in SMAP SM retrieval. The results of this study have shown a good agreement of retrieved SM when compared to in-situ SM from International Soil Moisture Network (ISMN). Especially, meaningful improvement was found in the dense forest regions compared to original SMAP SM retrievals (version 8 of SCA-V and DCA). The results of this study can help understand various SM-related hydrological processes.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2022
- Bibcode:
- 2022AGUFM.H22R1077L